Document Type
Presentation
Abstract
Developmental dysplasia of the hip (DDH) is described as under-coverage of the femoral head by the acetabulum, resulting in mechanical instability. Though DDH is often diagnosed using plain film radiographs, these images cannot adequately capture 3D joint coverage. Herein, we applied a 3D statistical shape model (SSM) to the femur and hemi-pelvis of patients with DDH to objectively measure shape variation and evaluated whether SSM outputs could predict measurements of joint coverage.
The femur and hemi-pelvis were semi-automatically segmented from CT images (83 hips from 47 females with DDH). Surfaces of each hip were reconstructed from segmentations, aligned, and input into a multi-domain SSM (shapeworks.sci.utah.edu). Correspondence particles were automatically placed over the bone surfaces and a subset on the femoral head and acetabulum were isolated for a joint-specific model. Modes of shape variation were determined with principal component analysis (PCA). A sparse model of PCA modes predicting coverage was determined using linear regression with Lasso regularization.
Coverage measurements ranged from 27.3% to 39.4%. Eight and 13 modes were selected for the full bone and joint-specific models, respectively. These modes represented 6.1% and 39.6% of the overall shape variation for full bone and joint-specific models with mean prediction errors of 0.9% and 0.6% coverage, respectively (Figure 1). Selected modes represented the depth of the acetabulum and oblateness of the femoral head, aligning well with the clinical description of DDH. In addition, the full bone model captured morphological and pose-related differences potentially related to altered muscle paths or differences in femoral torsion.
Included in
Biomechanics and Biotransport Commons, Musculoskeletal System Commons, Orthopedics Commons
Application of Statistical Shape Modeling to Predict Clinical Metric of Femoral Head Coverage in Patients with Developmental Dysplasia
Developmental dysplasia of the hip (DDH) is described as under-coverage of the femoral head by the acetabulum, resulting in mechanical instability. Though DDH is often diagnosed using plain film radiographs, these images cannot adequately capture 3D joint coverage. Herein, we applied a 3D statistical shape model (SSM) to the femur and hemi-pelvis of patients with DDH to objectively measure shape variation and evaluated whether SSM outputs could predict measurements of joint coverage.
The femur and hemi-pelvis were semi-automatically segmented from CT images (83 hips from 47 females with DDH). Surfaces of each hip were reconstructed from segmentations, aligned, and input into a multi-domain SSM (shapeworks.sci.utah.edu). Correspondence particles were automatically placed over the bone surfaces and a subset on the femoral head and acetabulum were isolated for a joint-specific model. Modes of shape variation were determined with principal component analysis (PCA). A sparse model of PCA modes predicting coverage was determined using linear regression with Lasso regularization.
Coverage measurements ranged from 27.3% to 39.4%. Eight and 13 modes were selected for the full bone and joint-specific models, respectively. These modes represented 6.1% and 39.6% of the overall shape variation for full bone and joint-specific models with mean prediction errors of 0.9% and 0.6% coverage, respectively (Figure 1). Selected modes represented the depth of the acetabulum and oblateness of the femoral head, aligning well with the clinical description of DDH. In addition, the full bone model captured morphological and pose-related differences potentially related to altered muscle paths or differences in femoral torsion.
Comments
shapeworks.sci.utah.edu